A milestone can protect a game by making it less risky

Production gates turn uncertainty into a schedule someone can fund. They also decide which uncertainty must be cut before it has a chance to become the game.

6 min · 14 sources cited

The tension

Game production's milestone gates meet industry economics at the funding decision that turns a risky creative possibility into an auditable deliverable.

What collides

Production control needs measurable milestones, certification readiness, and bounded scope, while creative discovery needs time for systems whose value cannot be known before they exist.

The question it replaces

Which kinds of uncertainty is a funded game permitted to keep?

A milestone can make a game safer by making part of the game less likely to exist.

That isn’t the usual story told about production gates.

They’re supposed to protect the project.

Turn a pitch into a prototype. A prototype into a vertical slice. A vertical slice into something a publisher can sensibly keep paying for.

The protection is real.

So is the subtraction.

Keep proving. Keep pricing. Keep converting a possibility into a deliverable with a payment date attached.

A prototype is allowed to be unruly.

It tests a mechanic in isolation. It needs no final art and no working save system, because it isn’t trying to be a game yet.

A vertical slice is a sterner object.

A short playable segment, near-final art, functional interface, the actual systems working together.

All at once. For a few minutes.

Between those two sit a first playable, an interim build, and eventually a vertical chunk.

Only then is anything called feature complete.

This isn’t bureaucracy pretending to be design.

It’s a way of finding out whether the pipeline can produce the game it has promised.

Before the budget is spent finding out the expensive way.

Full production then scales a validated pipeline.

More levels. More assets. More encounters.

The bulk of the money follows the proof.

Content is ruinously expensive exactly when the process for making it is still a rumour.

But a validated pipeline has a preference, and the preference is the problem.

It prefers what can be repeated. It prefers known quantities.

It prefers the system that survived the slice to the system that might only reveal its purpose after the slice has been dismantled and rebuilt.

Anthem gives that preference a grim calendar.

After a spectacular public reveal, reporting described a game still effectively in preproduction, with barely a mission built.

Roughly a year remained to compress the whole of full production into.

A gate can’t discover a production method after a team has spent the schedule pretending one exists.

It can only record the lateness more precisely.

That is the case for control.

Measurable milestones let a publisher ask whether a build contains the contracted features.

Whether it hits its stability bar. Whether it meets its stated gameplay goals.

A rejected build enters a cure period, then returns for resubmission.

The argument is contractual because the disagreement is contractual.

Done is not a mood, and neither is the payment.

The cost is that a good question can look identical to a bad deliverable.

A mechanic still being discovered can’t honestly claim feature completion.

A level whose purpose keeps changing under playtest evidence can’t promise last month’s content count.

Exploration is hard to audit.

It is, among other things, an admission that the answer hasn’t been found yet.

Accounts departments have rarely been moved by the beauty of an unresolved interaction.

The production vocabulary makes all of this sound tidier than it is.

Alpha is feature-complete. The planned systems work, even where content is still partial.

Beta is content-complete. The intended levels, assets and narrative material exist, and work turns to bug triage, balance, performance and localization.

The distinction preserves a place for unfinished content.

It preserves very little permission to discover a new system after alpha.

Not without making the word complete perform a small fraud.

Measurement isn’t one instrument waiting at the end of the schedule either.

Moderated playtesting puts a researcher with one player.

It survives an unstable build, because the researcher can intervene. It asks why.

Unmoderated sessions run in parallel against stable builds. They ask how many.

Telemetry records session length, progression speed, completion rates, ability use, resource consumption and matchmaking outcomes. It needs a population large enough to generalize from. It arrives last.

Each instrument protects a different fiction.

A milestone review can pass while the game plays badly.

A moderated session can find a confusion without measuring how often it happens.

A dashboard can show a drop-off, a death heatmap, an overused ability.

It explains no cause at all.

Telemetry answers what, not why.

The convenience of a graph is that it hands you a number.

The inconvenience is that the number has not agreed to explain itself.

Certification makes the outside world literal.

A platform holder’s process separates findings that block release from findings that only get recorded.

The first kind arrives with a resubmission calendar attached.

A failed final submission returns carrying every blocking issue.

Then travels through review again from the start.

The build cannot negotiate with a calendar after the calendar has become a rulebook.

There’s a clean economic reason for the severity.

A storefront takes its cut before the publisher sees anything.

A milestone-funded developer is paid against deliverables.

Not a running share of that sale.

So the publisher carries the sales risk.

The milestone exists to make the development risk legible before any sale arrives.

The arithmetic behind that is not subtle.

Take a large modern budget. Add a marketing spend of comparable size. Divide by what actually reaches the publisher per copy.

Break-even lands in the millions of units, before a single discount is applied.

The calculation ignores discounting, retailer margins, overhead and cost of capital.

It’s an order-of-magnitude model, not a forecast.

It’s still enough to explain why an unproven feature acquires the moral atmosphere of an unpaid invoice.

The economic case for the gate is strongest when it refuses romance.

A milestone schedule stops a team scaling a broken pipeline. It protects salaries from an indefinitely exploratory project. It forces whoever holds the money to say what they’re paying for.

It gives uncertainty a boundary and a price.

And a chance to be corrected before the cost becomes structural.

The creative case is strongest when it refuses its own romance too.

Discovery can be delay with better vocabulary, and often is.

Yet the system nobody could justify before building it is sometimes the one thing that makes a finished game more than its approved slice.

A production that cuts every unmeasured possibility stays fundable.

It can also stay merely competent.

Competent is not a thing anyone remembers.

The funded schedule needs credible evidence.

The game needs room for evidence that doesn’t exist yet.

One preserves a project by reducing uncertainty.

The other preserves a possibility by tolerating it.

Which is much harder to put on a schedule.

Neither side of that is unreasonable, which is why the argument never ends.

Which kinds of uncertainty is a funded game permitted to keep?

What it establishes

Preproduction tests individual mechanics, while a vertical slice proves the near-final pipeline before full production scales it.
Alpha is feature-complete and beta is content-complete, so the two gates describe different kinds of finish.
Moderated playtesting explains why players behave as they do, while unmoderated testing measures behavior across a larger stable panel.
Telemetry records what players do at scale but does not diagnose why a drop-off occurs.
Certification is a platform-holder review with submission checks, build verification testing, and Xbox Requirement testing.
A digital sale reaches the publisher only after the storefront applies its take rate, while physical retail follows a separate wholesale path.
Break-even arithmetic uses sold-through revenue at achieved prices and must not confuse shipped units with actual sales.

Asked and answered

What does a milestone actually buy in game development?

A milestone buys a reviewable claim about progress and usually releases a scheduled development payment when the claim is accepted. A publishing contract can define a prototype, First Playable, Vertical Slice, or Feature Complete as separate deliverables. The gate turns an uncertain production into a sequence a publisher can inspect, fund, cure after rejection, and ultimately stop funding if the contract's remedies are reached.

What is the difference between a prototype and a vertical slice?

A prototype tests an individual mechanic in isolation and can use placeholder art, incomplete systems, and disposable code. A vertical slice is a short fully playable segment with near-final art, functional UI, and the actual systems working together. The prototype asks whether a mechanic is promising; the vertical slice asks whether the production pipeline can make the promised game at finished quality.

Why do contracts split one game into so many milestones?

A real publishing-agreement schedule can contain twelve stages over roughly 2.5 years, including Prototype, First Playable, Interim Milestone, Vertical Slice, Vertical Chunk, Feature Complete, Release Candidate, First Party Certification, Game Release, and Final Build Archive. Each stage gives the funder a specific deliverable to inspect before releasing the next payment. The schedule converts progress into contractual evidence rather than relying on a general promise that development is advancing.

Is alpha the same thing as beta?

No. Alpha is feature-complete: every planned system and mechanic functions, although levels, assets, encounters, and narrative content can still be incomplete. Beta is content-complete: the intended levels, narrative content, and assets exist, and work shifts toward bug triage, balance, performance optimization, and localization. The labels describe different risks, not two interchangeable degrees of almost finished.

Why is a publisher milestone review more than ordinary QA?

A publisher milestone review tests a build against negotiated acceptance criteria such as feature completion, stability thresholds, performance targets, or specified gameplay goals. If the review rejects the deliverable, a contract commonly grants a cure period for correction and resubmission; some contracts also limit resubmission attempts before remedies including termination become available. The review is adversarial by design because payment and continuation depend on the result.

When does moderated playtesting tell a team more than unmoderated testing?

Moderated playtesting places a researcher with one player in real time. The researcher can ask follow-up questions, use debug controls, route around a broken build, and identify why a reaction occurs, so this method works in preproduction and early full production. Unmoderated testing runs many self-recorded sessions in parallel, which measures frequency and balance more efficiently once a build is stable without researcher intervention.

What can telemetry tell a game team that a playtest cannot?

Telemetry automatically records behavior across a broad beta or soft-launch population, including session length, progression speed, completion rate, weapon or ability use, accuracy, damage, resource consumption, and matchmaking outcomes. Dashboards can expose a death heatmap or a progression drop-off at a precise point. Telemetry establishes what players do and how often; it does not establish why they do it, which still requires qualitative investigation.

What happens during Xbox certification?

Microsoft's documented Xbox process begins with submission checks, including Submission Validator and malware scanning. It then runs build verification testing, or BVTs, to confirm that the build is testable before Xbox Requirement, or XR, testing examines the applicable requirements. A Standard Reporting Issue does not block release, a Condition for Resubmission does, and an Issue of Note is optional to fix because it is not a formal requirement.

How much time can a failed certification pass cost?

Microsoft publishes five business days for a disc submission and four business days for a digital base final submission; a digital Fastlane submission can take as little as 48 hours. Standard content updates run two business days and Fastlane updates 24 hours. A final submission that receives a Condition for Resubmission must be corrected and sent through the process again, so failure adds another review cycle rather than a simple patch-and-retest.

Where does the money from a digital game sale go first?

The storefront deducts its take rate from the list price before remitting the balance to the publisher. In the corpus's fictional $69.99 console-storefront trace, a 30% cut retains roughly $21.00 and leaves approximately $48.99 for the publisher. A milestone-funded developer is generally paid against agreed deliverables rather than a percentage of every sale; the publisher carries sales risk and usually keeps the post-platform remittance.

Why is a 30 percent storefront cut not a universal rule?

A 2019 IGN survey found a 30% baseline across Steam, GOG, the Microsoft Store, Nintendo Switch, and both major consoles of that generation, but storefront terms vary. Epic Games Store has used a 12% rate since its 2018 launch, which would leave about $61.59 from the same fictional $69.99 sale. Steam also uses 30%, 25%, and 20% revenue bands at different thresholds. The storefront changes the publisher's remittance.

Does break-even mean a game has shipped enough copies?

No. Break-even concerns revenue actually received against total costs, not copies shipped into a channel. In a simplified full-price digital trace, Cyberpunk 2077's disclosed $174 million development cost and $142 million marketing cost, $316 million combined, divide by $48.99 to about 6.5 million units. Discounting, retailer margin, publisher overhead, cost of capital, and achieved prices make real break-even less simple than that illustration.

Why is physical retail not just digital sales with a different platform cut?

A digital sale moves from platform holder to publisher after the storefront take rate. A physical sale adds a retailer that buys inventory from the publisher at a wholesale price below the shelf price and keeps the difference. The platform holder still licenses and approves manufacturing, but does not clear the retail payment in the same way. The two sales paths therefore contain different intermediaries and different margins.

Game design in the AI era

Bring a sharper eye to your next game.

Making games while the tools keep changing? Get our research brief on AI, game design and production. See what generated assets and faster answers mean for the work—and bring better questions to your next design discussion.

Occasional dispatches, when there is something worth your attention.